I've observed the same, so many forks of forks of forks happening right now rather than group efforts. But I think this is coming from the fact that PR review is incredibly costly to the maintainer while generating code is fast. I've submitted PRs to open source projects which claim to welcome contributions but the PRs just never get reviewed (this was an issue pre AI too). If you wanted to build something, constantly fighting an uphill battle to get PRs merged is less appealing than just creating your own project where you can work full speed on whatever you want.
I'm not too sure what the future here looks like and if this is even a problem, but things certainly are changing. These days submitting a well written bug report or feature suggestion, or validating existing reports is more useful than sending a PR.
maybe the focus should shift a bit to using AI for PR reviews. adversarial reviews, summarizing it for a human. we'll see where time takes us...